Background: Amyloid imaging data are often quantified using a method where target region to cerebellar cortex tracer uptake ratios (SUVRcer) are computed. We investigated the agreement between blinded visual evaluations (BIE) and quantitative analysis of [18 F]f lutemetamol in a Japanese population.
      Methods: Healthy volunteers (HV; n=25, 15 elderly healthy volunteer ≥ 55 years (EHV), 10 young healthy volunteers <55 (YHV)) and patients with probable Alzheimer's Disease (pAD; n=20) and amnestic mild cognitive impairment (aMCI; n=20) each underwent a 30min dynamic PET scan, starting 90min post injection of 185MBq [18 F]flutemetamol. A 3D T 1 MRI was also obtained. Image data was processed to allow for application of a template set of regions of interest to a 30 min summation image to estimate a global cortical average SUVRcer. The optimal threshold was estimated as the midpoint between the mean of pAD and EHV cohorts in terms of standard deviation for the global SUVRcer. BIE were performed by 5 Japanese and 5 non-Japanese independent board-certified readers to obtain the majority outcomes of the ten readers. T he diagnostic capability of BIE and of global SUVRcer for differentiating AD from HV was evaluated, together with agreement between BIE and global SUVRcer.
      Results: Optimal threshold for discriminating pAD from HV was 1.391 for the global cortical SUVRcer average. Sensitivity and specificity of discriminating pAD from HV in this cohort for BIE (majority outcome) were 90% and 100%, respectively, and for SUVR cer 90% and 100%. In the efficacy evaluation, the quantitative classification categorized 18 subjects as positive (18 pAD, 0 HV), all of these were classified as positive by BIE. 27 subjects were categorized as negative quantitatively (25 HV, 2 pAD) and also by BIE, resulting in a 100% agreement between BIE and quantification using this threshold. BIE categorized 10 of 20 aMCI subjects as positive, quantitative classification added three subjects to this category.
      Conclusions: Agreement between the visual and quantitative assessment of [18 F]flutemetamol images is high. While quantitative classification should not replace visual assessment, it may be useful as a tool for detecting early stages of amyloid accumulation.